Re: [Chicken-users] Posix threading

[Julian Morrison]

Toby Butzon wrote:

> I'm new to "real" Scheme programming[1], but I'm quite surprised
> that it doesn't use posix/native threading.
Some of the interpreted schemes probably do. I haven't looked. Posix 
threading with dynamic code is perfectly possible, I know that java and 
ocaml both use Posix threads.

> It seems like a wacky
> idea to have threads that can't be scheduled on separate CPUs, or
> worse, that block the entire process on a blocking system call
> (surprising since this is one of the major ways threads are used
> in other languages).
>  
Better than not having any threads. Many blocking kernel calls can be 
emulated via nonblocking IO. (This obviously fails when calling out to 
external code which does not use scheme's IO primitives.)

> Apologies in advance if these are stupid questions, but what am I
> missing?
Chicken's GC mechanism.

If I read the docs right, what it does is approximately:
- convert function calls from call-and-return to continuation-passing-style
- allocate all objects on stack
- when the stack gets too big, copy all reachable stack objects to heap, 
and unwind the stack
- periodically do a copying GC in heap

Problems:
- Posix threads have seperate stacks. Stack allocated objects in one 
thread are unreachable from others.
- Copying GC moves pointers in heap. If one thread does a copying GC, 
the others won't know where the pointers moved.

Possible solutions, respectively:
- Do a minor GC as though stack space had run out, before starting each 
Posix thread.
- Either
--- access pointers indirectly through some means which makes their 
movement in heap transparent (evicted locatives?),
--- or provide a mechanism to temporarily lock current (to be shared) 
objects against GC, which is called before any threading.


> Is it just that scheme semantics make it hard?
>  
More like, continuations make userspace threads easy.